The Friction Factories
Two decision types. Three friction types. Four consultancy archetypes. Yet everyone is selling the same solution.
TL;DR
Every major consulting firm sells the same AI transformation framework because clients have always rewarded friction removal over risk architecture.
But there are three kinds of friction: cognitive, operational and accountability.
The first two are two-way doors - you can rebuild them if they break.
The third is a one-way door.Each consulting archetype attacks a different friction type:
MBB removes cognitive friction,
integrators remove operational friction,
specialists protect accountability friction,
boutiques are the only archetype built to add it back.
The 56% of CEOs getting nothing from AI is the direct output of firms that removed the wrong friction from the wrong decisions.
This article maps which archetype to hire for which decision type.
And what happens to your risk architecture when you get it wrong.
Last week (April 9), Bernard Huber published The Big Consulting AI Frameworks, Compared - a rigorous side-by-side of how McKinsey, BCG, Deloitte, EY, PwC, KPMG, Accenture, Bain, Capgemini and IBM frame AI transformation in 2026.
It is well-sourced, honest about the gap between branding and substance. And it ends with three sharp buyer questions most RFP processes never ask. Read it first.
I will take this one step further.
After reading Huber’s analysis, I wanted to answer four questions the comparison leaves open:
Why do all ten frameworks land on the same four workstreams?
What structural force makes every firm sell essentially the same thing?
Which archetype should your company actually hire, and for which type of decision?
What happens to your organization when you hire the wrong one?
I find the main symptom in two numbers:
PwC’s 29th Global CEO Survey: only 12% of CEOs say AI has delivered both cost and revenue benefits. 56% say they are getting nothing out of it.
Deloitte’s State of AI in the Enterprise 2026: only 1 in 5 companies has a mature governance model for autonomous AI agents.
Here’s my diagnosis.
Why every framework looks the same
Ten firms. Ten different brand narratives. One identical structure underneath: strategy and use-case selection, data and infrastructure, deployment and change management, governance and risk.
This similarity signals something more than a mere coincidence.
Clients buy friction removal. They don’t buy outcomes, nor governance, nor risk architecture. Just friction removal. Executives measure consulting ROI in hours saved, headcount reduced and cycle times compressed. Every firm built its framework to win that RFP. The four workstreams mirror the procurement requirements.
And that is why 56% of CEOs are getting nothing. They bought friction removal. But the friction they removed was the wrong kind.
The three frictions
I wrote about this in detail here. The short version:
Cognitive friction is the mental drag that forces you to think before making a consequential decision. AI removes it by summarizing, pre-selecting and auto-completing.
Operational friction is the checklists, the two-manager sign-offs, the manual reconciliations. The shock absorbers between a bad input and a catastrophic output. AI bypasses them in seconds.
Accountability friction is the audit logs, the explainability requirements, the human-in-the-loop review. The layer that the EU AI Act now mandates for high-risk systems with fines up to 7% of global revenue for organizations that skip it.
Here is the decision rule:
Removing cognitive or operational friction is usually a two-way door. The process breaks, you put the checklist back. Painful but reversible.
Removing accountability friction is a one-way door. Once you are audited, breached, or cited in litigation for an ungoverned agent failure, the event has happened. You cannot un-trigger it.
Every consulting archetype attacks one of these three frictions. AI made each of them faster at it. What nobody tells you is which attack is safe.
The four consulting archetypes
I analyse each archetype on two levels:
what they must change internally to stay aligned with the AI era,
and where they need to help clients reduce or increase friction by decision type.
1. The strategy factories - MBB
McKinsey, BCG and Bain (MBB) are in the business of removing cognitive friction for the C-suite.
What MBB must change internally
McKinsey now runs 25,000 AI agents alongside 40,000 humans, with AI compressing the research and options-generation work that once took analyst teams weeks. That is a two-way door for their internal operations. Reversible if quality drops, and the senior layer still catches errors.
The internal one-way door is their career ladder. If they eliminate the broad junior intake permanently, replacing apprenticeship with AI, they destroy the pipeline that produces future senior judgment.
You can re-hire humans. You cannot re-grow the institutional instinct that comes from years of client exposure. Internally, MBB must treat the talent model as a one-way door. And keep deliberate friction in any decision that changes it permanently.
Where they must help clients reduce or increase friction
MBB should reduce cognitive friction for two-way door decisions: exploratory analysis, option generation, scenario modelling. That is where speed and synthesis genuinely help.
They should actively add cognitive friction for one-way door decisions, like board-level restructurings, vendor lock-ins, irreversible operating model changes. Force the board to interrogate the AI’s logic before the door closes, not after. It shouldn’t be a consulting add-on. It must become the core service, stripped of the part the agent already does better.
2. The integrators - Big 4 and IT services
Deloitte, PwC, Accenture and Capgemini are built to remove operational friction at industrial scale: automated workflows, offshored back-office tasks, standardized ERP implementations.
What integrators must change internally
PwC declared the pyramid dead in January 2026, predicting AI agents filling the execution middle of every major organization. Internally, the Big 4 are living this already. They have leaner junior intake, AI handling delivery volume, partners selling outcomes rather than hours.
The internal one-way door is standardization speed. When a delivery pattern works on one client, the temptation is to codify it into a global template and scale it immediately. Once that template is deployed across hundreds of clients in regulated industries, reversing a design flaw is not a project. It is a sector-wide recall.
Internally, integrators need more friction - slower gates, independent review - before any pattern becomes a global method.
Where they must help clients reduce or increase friction
Reduce operational friction aggressively for low-risk, two-way door processes: invoice routing, IT ticket triage, HR onboarding, anything with clear rollback paths and limited regulatory exposure.
Add operational friction - explicit checkpoints, human review steps, kill switches - for high-risk, one-way door transformations: core financial systems, identity and access management, safety-critical processes.
The 1-in-5 governance maturity number is what happens when integrators only deliver the first half.
3. The global specialists - domain authorities
Firms like Oliver Wyman and Alvarez & Marsal do not sell speed. They sell credibility: the risk models, regulatory frameworks, and domain expertise that give banks and healthcare companies their operating license.
What specialists must change internally
Internally, the temptation is to embed AI into the models to generate scenarios faster and compete on speed.
That is the most dangerous internal one-way door on this list.
If AI quietly shapes the authoritative methodology and the regulator later cannot trace the logic, the firm's credibility built over decades is damaged in a single audit cycle.
Internally, specialists need to treat methodology changes as one-way doors: high accountability friction, external validation, full documentation before any AI component enters a client-facing model.
Where they must help clients reduce or increase friction
Reduce cognitive friction for client-facing complexity: use AI to make risk and compliance understandable without stripping the nuance.
That translation is genuinely useful. Protect and strengthen accountability friction for every regulated, one-way door decision: capital adequacy models, safety submissions, medical device approvals.
The specialist’s job is not to make compliance faster like Delve tried. It is to make compliance defensible when the regulator asks who made the decision and how.
4. The boutiques - surgical teams
Small, senior-heavy boutiques with typically 50 to 300 people, and with narrow domain focus. They cannot compete on factory-scale friction removal. Their historic strength is context, senior attention and trust.
What boutiques must change internally
The internal one-way door for boutiques is scope creep disguised as growth: taking on large-scale transformation work because the AI tools make it look deliverable.
One failed implementation at factory scale can eliminate the trust that took years to build.
Internally, boutiques need social friction around engagements that fall outside their surgical model - a deliberate, high-threshold process for saying no to work that requires scale they do not have.
Where they must help clients reduce or increase friction
Reduce cognitive friction at the board and executive level: translate technical AI risk into the three languages boards actually decide in - fiduciary exposure, financial consequence, competitive precedent. That translation is invisible on a cost-per-slide comparison and irreplaceable in a room where a one-way door is being discussed.
Build accountability friction for every one-way door deployment: design the decision logs, name the owners, define what failure looks like before the agent goes live.
Boutiques are the only archetype with no factory revenue to protect and no platform to sell. Their independence is the product.
What happens when you hire the wrong archetype
The dangerous mismatches always follow the same pattern: the friction being removed does not match the reversibility of the decision being made.
Hiring a strategy factory to validate a one-way door gives you a flawless slide that makes the irreversible look inevitable. The cognitive friction that should have slowed the room down is gone.
Hiring an integrator to automate a high-risk regulated process gives you a globally deployed pattern with no shock absorbers. The operational friction that was supposed to catch the edge case is gone.
Hiring a specialist that has quietly embedded AI into its methodology gives you authoritative-looking compliance work with an accountability layer nobody can inspect.
Hiring a boutique to do factory-scale implementation gives you senior attention on work that needs throughput, and a burned relationship when the volume breaks the model.
In every case:
wrong archetype or wrong friction type = wrong decision.
Yes, we could argue clients should know what they are buying. And they should.
But not one of the ten firms in Huber’s comparison will walk into a pitch and say: “The work you are asking us to do will remove the only friction standing between you and a bad one-way decision.”
That conversation kills the deal. So it never happens.
Which archetype do you actually need?
Two questions. Answer both before you issue the RFP.
What type of friction does this decision require you to manage?
Is the decision reversible, or is it a one-way door?
Your answers determine the archetype. Everything else - brand, framework quality, slide aesthetics - is irrelevant.
You need a strategy factory (MBB) when the decision is genuinely complex, the options space is large, and the cognitive load of structuring the problem is the bottleneck.
The right use is two-way door exploration: scenarios, trade-offs, option generation.
The wrong use is asking them to validate a decision you have already made. That is not strategy. That is expensive cognitive friction removal on a one-way door that is already closing.
You need an integrator (Big 4 / IT services) when the decision is already made and the job is execution at scale.
Operational friction removal across defined processes, with a clear rollback path if something breaks.
The wrong use is asking them to govern what they are building. Their incentive is throughput. Governance slows throughput. Do not put the same firm on both sides of that tension.
You need a global specialist when the decision lives inside a regulated domain and the accountability friction is critical. Risk models, compliance frameworks, sector-specific methodologies.
The right use is situations where being wrong carries license-level consequences.
The wrong use is expecting them to move fast. If they are moving fast, they are cutting the friction that justifies their existence.
You need a boutique when the decision is a one-way door and the stakes are high enough that you need someone in the room whose only job is to make you think harder before you walk through it.
Decision architecture, board-level translation, accountability design.
The wrong use is scale. If you need 200 people and six months of implementation, a boutique is not the answer.
You need to build it internally when the capability is core to how your organization will make decisions for the next five years.
Hiring any external archetype to do work that belongs on your permanent capability map is operational friction removal dressed as strategy.
You are outsourcing a muscle you will need permanently.
The consultants leave. The one-way door stays.
My final ask
These Signals reflect conversations I am having with executives right now, just written down.
If this helped you see which friction your consultants are actually removing - and whether that is safe - do two things.
Key Sources
The Big Consulting AI Frameworks, Compared, No More Pyramids: Rethinking Your Workforce for the Agentic AI Era, McKinsey Now Has 60,000 Employees: 25,000 of Them Are AI Agents, PwC 29th Annual Global CEO Survey, One-Way Doors and AI Strategy, AI readiness cult.
Thanks for sharing this post and it’s an interesting read.
I do have a few thoughts. A lot of the issues that large corporations that buy consulting services have are due to underlying business models, cultures and defaults.
For a long time now, the accepted logic has been “bigger is better” and all consulting mainly focuses on helping already large businesses get larger and reducing friction in that larger machine from the perspective of leadership.
However, for a while now (and they excitement about LLMs just made it more obvious) there are many areas where bigger is much much worse.
Most consulting AI projects fail because it’s not addressing the right underlying problems. Exactly to your point they are trying to remove more friction for leadership in an op model where less friction doesn’t help anyone.
I liked the article. Just feel the tldr doesn’t do justice to the entire piece.